I've sometimes used in Alexandroff one-point compactification in scenarios where you really "need the zero to have an inverse" and it works well in certain applications.